Diagnosing Autism Spectrum Disorder Using Natural Language Processing

1. A system, comprising: a computer processor; and a memory containing a program that, when executed on the computer processor, performs an operation for processing data, comprising: generating a machine learning (ML) model using training data comprising a first plurality of training examples, each example being a text of a conversation labeled as exhibiting at least one characteristic of autism; receiving text of an ongoing conversation between a plurality of participants; annotating the text of the conversation using natural language processing; identifying features in the conversation using the annotations; evaluating the features using the ML model to determine a measure of probability that a first one of the plurality of participants in the conversation falls on the autism spectrum; and based on the measure of probability, outputting for display, during the conversation, a notice indicating that the first participant has exhibited a characteristic of autism.

2. The system of claim 1 , wherein the training data comprises a second plurality of training examples, each example being a text of a conversation labeled as not exhibiting any characteristic of autism, wherein the ML model is generated using both the conversations labeled as exhibiting at least characteristic of autism and the conversations labeled as not exhibiting any characteristics of autism.

3. The system of claim 2 , wherein generating the ML model comprises: identifying features in the conversations in the first and second pluralities of training examples using natural language processing, each of the features describing at least one of a pattern and a behavior in the conversations.

4. The system of claim 3 , wherein generating the ML model comprises: assigning a respective weights to the features in the ML model.

5. The system of claim 1 , wherein one of the features describes at least one of a pattern and a behavior in the textual conversation and corresponds to at least one characteristic of autism.

6. The system of claim 5 , wherein evaluating the features with the ML model comprises: assigning weights to the plurality of features, wherein the weights represents a likelihood that a corresponding feature is a characteristic of autism.

7. A computer program product for diagnosing autism, the computer program product comprising: a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to: generate a ML model using training data comprising a first plurality of training examples, each example being a text of a conversation labeled as exhibiting at least one characteristic of autism; receive text of an ongoing conversation between a plurality of participants; annotate the text of the conversation using natural language processing; identify features in the conversation using the annotations; evaluate the features using the ML model to determine a measure of probability that a first one of the plurality of participants in the conversation falls on the autism spectrum; and based on the measure of probability, output for display, during the conversation, a notice indicating that the first participant has exhibited a characteristic of autism.

8. The computer program product of claim 7 , wherein the training data comprises a second plurality of training examples, each example being a text of a conversation labeled as not exhibiting any characteristic of autism, wherein the ML model is generated using both the conversations labeled as exhibiting at least characteristic of autism and the conversations labeled as not exhibiting any characteristics of autism.

9. The computer program product of claim 8 , wherein generating the ML model comprises computer-readable program code further executable to: identify features in the conversations in the first and second pluralities of training examples using natural language processing, each of the features describing at least one of a pattern and a behavior in the conversations.

10. The computer program product of claim 9 , wherein generating the ML model comprises computer-readable program code further executable to: assign a respective weights to the features in the ML model.

11. The computer program product of claim 7 , wherein one of the features describes at least one of a pattern and a behavior in the textual conversation and corresponds to at least one characteristic of autism.

12. The computer program product of claim 11 , wherein evaluating the features with the ML model comprises computer-readable program code further executable to: assign weights to the plurality of features, wherein the weights represents a likelihood that a corresponding feature is a characteristic of autism.